A combination of support vector regression, structure-based pharmacophore and virtual screening for cytochrome P450 2C9 inhibitors from Chinese herbs

Abstract

Cytochrome P450 2C9 (CYP2C9), one important isoform of the cytochrome P450 (CYPs), mediates the oxidation of some important drugs. The inhibitors of this target often affect the metabolic rate of the corresponding metabolites and then result in undesirable drug-drug interactions (DDIs) in clinical. In order to discover potential CYP2C9 inhibitors, a support vector regression (SVR) model with good predictive ability were constructed. The correlation coefficient (R 2) and mean square error (MSE) values of the optimal SVR model were 0.952 and 0.003. Meanwhile, a structure-based pharmacophore (SBP) model was generated based on the crystal complex of CYP2C9 (PDB ID: 4NZ2) to refine the results of SVR model and elucidate the mechanism of the inhibitors. The best SBP model consists of one hydrogen bond acceptor feature, three hydrophobes features and six exclusion volumes. Then, both the optimal models of SVR and SBP were utilized to predict compounds in Traditional Chinese Medicines Database (TCMD) to identify potential CYP2C9 inhibitors from Chinese herbs. Finally, 1514 compounds were reserved, whose predicted active values obtained from SVR model and Fitvalues obtained from SBP model were all higher than the corresponding values of the initial compound in 4NZ2. Among them, ID 14767, which has higher predicted values and better mapping results with the SBP model, might exhibit inhibition effect on CYP2C9. Both the SVR model and SBP model might be applied in discovering potential CYP2C9 inhibitors from Chinese herbs, and also provide reference for the rational application of drugs in clinical.